ddlau / papers

PART I

Base

• Playing Atari with Deep Reinforcement Learning
• Human-level control through deep reinforcement learning
• Rainbow: Combining Improvements in Deep Reinforcement Learning

Double

• Deep Reinforcement Learning with Double Q-learning (double)

Dueling

• Dueling Network Architectures for Deep Reinforcement Learning (dueling)

Manipulate Experience Replay

• PRIORITIZED EXPERIENCE REPLAY (PER)
• Prioritized Sequence Experience Replay (PSER)
• Hindsight Experience Replay

Distributed

• Massively Parallel Methods for Deep Reinforcement Learning (Gorila)
• Asynchronous Methods for Deep Reinforcement Learning (A3C)
• REINFORCEMENT LEARNING THROUGH ASYNCHRONOUS ADVANTAGE ACTOR-CRITIC ON A GPU (GA3C)
• EFFICIENT PARALLEL METHODS FOR DEEP REINFORCEMENT LEARNING (PAAC)
• ACCELERATED METHODS FOR DEEP REINFORCEMENT LEARNING
• IMPALA: Scalable Distributed Deep-RL with ImportanceWeighted Actor-Learner Architectures (IMPALA)

Distributional

• Statistical Aspects of Wasserstein Distances
• The Cramer Distance as a Solution to Biased Wasserstein Gradients
• A Distributional Perspective on Reinforcement Learning (C51)
• An Analysis of Categorical Distributional Reinforcement Learning
• Distributional Reinforcement Learning with Quantile Regression (QR-DQN)
• Implicit Quantile Networks for Distributional Reinforcement Learning (IQN)
• Fully Parameterized Quantile Function for Distributional Reinforcement Learning (FQF)
• Statistics and Samples in Distributional Reinforcement Learning (ER-DQN)
• A distributional code for value in dopamine-based reinforcement learning
• QUOTA: The Quantile Option Architecture for Reinforcement Learning

Exploration

• Deep Exploration via Bootstrapped DQN
• Parameter Space Noise for Exploration
• Noisy Networks for Exploration

Unclassified

• Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor

PART II

DPG

• Deterministic Policy Gradient Algorithms
• CONTINUOUS CONTROL WITH DEEP REINFORCEMENT LEARNING
• Distributed Distributional Deterministic Policy Gradients1

FIM & NGD & TRM

• Natural Gradient Works Efficiently in Learning 1
• A Natural Policy Gradient 1
• Approximately Optimal Approximate Reinforcement Learning 1
• New insights and perspectives on the natural gradient method
• Trust Region Policy Optimization
• High-Dimensional Continuous Control Using Generalized Advantage Estimation
• Generative Adversarial Imitation Learning
• Sample Efficient Actor-Critic with Experience Replay
• Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
• TensorFlow Agents: Efficient Batched Reinforcement Learning in TensorFlow
• Proximal Policy Optimization Algorithms
• Emergence of Locomotion Behaviours in Rich Environments
• Implementation Matters in Deep Policy Gradients: A Case Study on PPO and TRPO

FLANKS

Data/Image Augmentation

• CutOut - Improved Regularization of Convolutional Neural Networks with Cutout

NORMALIZATION

• BN - Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• LN - Layer Normalization
• IN - Instance Normalization: The Missing Ingredient for Fast Stylization
• GN - Group Normalization
• BIN - Batch-Instance Normalization for Adaptively Style-Invariant Neural Networks
• BRN - Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models
• CBN - Cross-Iteration Batch Normalization

ATTENTION

• SE - Squeeze-and-Excitation Networks
• CBAM - CBAM: Convolutional Block Attention Module

CNN

• AlexNet - ImageNet Classification with Deep Convolutional Neural Networks
• ZFNet - Visualizing and Understanding Convolutional Networks
• dropout - Improving neural networks by preventing co-adaptation of feature detectors
• maxout - Maxout Networks
• Network In Network
• VGG - Very Deep Convolutional Networks for Large-Scale Image Recognition

BACKBONES

• CSP - CSPNet: A New Backbone that can Enhance Learning Capability of CNN

NECKS

• SPP - Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
• FPN - Feature Pyramid Networks for Object Detection
• PAN - Path Aggregation Network for Instance Segmentation

OBJECT DETECTION

Pose Estimation

• Pose Machines: Articulated Pose Estimation via Inference Machines
• Convolutional Pose Machines
• Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
• OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
• PARE: Part Attention Regressor for 3D Human Body Estimation